Storage Tiering

1
Storage Tieringassigning the most reliable,
high-performance equipment to support the most
critical data, and the most cost-effective
resources to support older, less critical
information

• Ken Walters
• PBS

2
Questions

• How many of you have a SAN?
• If you dont, do you think it would be too
  expensive complicated?
• If you do, are you using a single large array?
• Are you considering iSCSI?

3
Agenda

• Definitions
• Justifications
• Background
• Tiered Model
• Enabling Technology
• Deployment

4
Definitions

• ATA/IDE (PATA)
• SATA (Serial ATA)
• SAS (Serial Attached SCSI)
• FC (Serial SCSI)
• SMB (Small/Medium Business)
• SAN (Storage Area Network)
• DAS (Direct Attached Storage)
• Virtualization

5
Justifications

• Lower disk costs
• SATA vs. FC
• Lower connectivity costs
• IP vs. FC
• Faster provisioning
• CAT5 vs. FC, iSCSI vs. FC
• Improved Productivity
• Foundation for ILM and Utility Computing

6
Background

• Concept has been around for a long time
• Mainframe environment
• Tape often a major component
• Large amounts of data
• Large amounts of money
• Not within reach of the SMB

7
Background

• Changing Times
• More being stored
• Retention Periods
• Disaster Recovery
• Logging
• Data mining
• Regulations - HIPPA, SOX

8
Background

• Typical SMB Data Center
• Technologies not conducive to efficient storage
  tiering
• Direct Attached Storage
• Single array
• Islands of storage
• NAS
• DAS
• SAN

9
Background

• Typical SMB Data Center
• Many-hat staffing
• No Storage Administrator
• DAS mindset
• Distributed mindset
• Many servers each running one thing
• Makes storage networking expensive

10
Tiered Model

• Storage is heavily centralized and networked
• Storage is well virtualized
• Storage arrays
• In-band appliances
• Gateways
• Server side virtualization

11
Tiered Model

• Multiple levels of storage
• Differing performance
• Differing availability
• Differing Service Objectives
• Problem resolution
• Backup and restore
• Recovery Time Objective (RTO)/Recovery Point
  Objective (RPO)

12
Enabling Technologies

• IP Storage
• Gateways
• Inexpensive disks
• Inexpensive arrays
• Multi disk type arrays
• Sophisticated Array Software
• Storage friendly PNP operating systems
• Low cost SAN components
• Zero Server Footprint arrays

13
Deployment

• If you dont have a SAN, build one and do it with
  iSCSI
• Use FC only where you need it and make sure you
  really need it.
• If you have a FC SAN, leverage it
• Gateway iSCSI
• Gateway NAS
• Consolidate Storage Islands

14
Deployment

• Replace equipment
• Arrays that support FC and SATA under common
  controllers (soon there will be SAS as well)
• Software to migrate between tiers
• Augment by adding a new array to your SAN
• SATA instead of FC
• Centralize Storage and attach as many servers as
  possible to drive down TCO
• Start deploying multiple tiers

15
Time Check

• Still to go
• Defining Tiers
• Defining Availability
• Implementing
• Benefits

16
Maple Syrup

• Grade A
• Grade B?
• Grade C?

• Grade A Light Amber
• Grade A Medium Amber
• Grade A Dark Amber

17
Defining Tiers
  Availability Performance Uses
AAA 99.99 High OLTP, ERP, Decision Support, Messaging, Databases
AA 99.9 High/Medium File Serving, Development, DR Mirrors, D2D Backup
A 99.8 Medium/Low Archiving, D2D Backup, Bulk Storage

• Avoid tier names that may disappoint
• Determine the attributes based on solid business
  analysis

18
AAA - Defined
Availability 99.99
Performance High write and read performance, rapid rebuild after disk failure
Disk Type FC, 15K, 73GB, 100 Duty Cycle
19
AA - Defined
Availability 99.9
Performance Read performance high, write moderate, rebuild time moderate
Disk Type FC/SATA, 15/7.2K, 73/250GB, 100 Duty Cycle
20
A - Defined
Availability 99.8
Performance Write performance low, read moderate, RAID5 rebuilds painfully slow. RAID0 may be a better choice, but no protection other than backups.
Disk Type SATA, 7,200/5,400K, 250/320GB, Duty Cycle 30/50?/100?

• Consider RAID0 if you choose very large disks

21
Measuring Availability
Uptime Downtime Downtime per Year Downtime per week
98 2 7.3 days 3 hrs 22 min
99 1 3.65 days 1 hr 41 min
99.8 0.2 17 hrs 30 min 20 min 10sec
99.9 0.10 8 hrs 45 min 10 min 5 sec
99.99 0.01 52.5 min 1 min
99.999 0.001 5.25 min 6 sec

• Avoid targets in italics
• People laugh at the top two now
• You lose sleep with the bottom one
• Cost of adding a 9 can be exponential

22
Availability Service Levels

• Defining availability targets is crucial
• Management approval
• Everyone wants five nines until they see the
  price tag. The operational costs can be
  staggering.
• Charge backs help people be realistic
• Use a Service Level Objective (SLO) not an
  Agreements

23
Implementing Tiering

• Sell the idea to management
• Classify your data
• Importance, security, lifespan, availability
• Mission Critical, Business Critical, Operational

24
Implementing Tiering

• Tier applications
• Attributes include
• Performance
• Availability
• Recoverability
• Security
• Disaster recovery RPO/RTO
• Define storage tiers
• Define server tiers

25
Benefits of Tiering

• Highest availability does not become your lowest
  common denominator.
• Storage costs are in sync with value of data
• Concentrate money and staff where it is needed
• Staff Productivity
• Tiered Storage is the foundation for
• Information Lifecycle Mgmt (90 process/10
  technology)
• Utility Computing

26
In Summary

• Build a SAN if you dont have one
• Use technology that supports tiering
• Tier your storage and applications
• Set availability service levels
• Sit back, save money and increase productivity

27
Resources

• Web
• snia.org/education
• storagenetworking.org
• searchstorage.techtarget.com
• google.com
• Books
• Building Storage Networks Marc Farley
• Resilient Storage Networks Greg P. Schulz

28
Thank You!

• kwalters_at_pbs.org